Maximal monotone operators with non-empty domain interior; characterizations and continuity properties M.D. Voisei Abstract In the context of general Banach spaces characterizations for the max- imal monotonicity of operators with non-empty domain interior as well as stronger continuity properties for such operators are provided. 1 Introduction In the context of locally convex or Banach spaces there are few characterizations for the maximal monotonicity of an operator (see e.g. [10, Theorem 3.8], [22, Theorem 2.3], and [26, Theorem 6]). All these characterizations are based on special convex representations associated to the operator. In a finite-dimensional space a complete characterization for the maximality of a monotone operator is given in [13, Theorem 3.4] in terms of direct oper- ator notions: the near convexity of its domain, convexity of its values, graph closedness, and behavior at the boundary of its domain. In a Banach space context characterizations of maximality, similar to those found in the finite-dimensional case, are available for full-space or open convex domain monotone multi-functions (see [4, Theorem 1.2], [8, Lemma 2.2], [21, Theorem 40.2, p. 155], [24, Lemma 4.2], and [17, Lemma 7.7, p. 104]). Every maximal monotone operator in a finite dimensional space has a non- empty convex relative interior of its domain which is dense in the domain and arXiv:1102.5055v1 [math.FA] 24 Feb 2011 a convex domain closure (see e.g. [15], [19, Theorems 6.2, 6.3], [20, Theorem 12.41, p. 554]). That is why, characterizations of the maximality of a monotone operator with non-empty relative (algebraic) domain interior in a general Banach space, in terms of notions directly linked to the operator and similar to those seen in the finite-dimensional case, constitute generalizations of all fore-mentioned results and that is our primary goal. Our secondary goal is to reveal several continuity properties with respect to the strong×weak-star topology for maximal monotone operators which have non-empty domain interiors and are defined in a normed barrelled space, such as, the closedness of the graph, the upper semicontinuity, and the Cesari property. 1 The plan of the paper is as follows. The next section introduces the reader to the main notions and notations used in this article. Section 3 dwells with the restrictions of an operator to affine subsets. Section 4 analyzes the finite- dimensional case and provides a new proof of [13, Theorem 3.4]. In Section 5 the characterizations previously seen in the finite dimensional context are extended to arbitrary Banach spaces via a hemicontinuity condition, demiclosedness, or representability. Section 6 deals with the continuity properties of monotone demiclosed operators that have a non-empty domain interior. 2 Preliminaries Let (E,µ) be a locally convex space and A ⊂ E. We denote by “conv A” the convex hull of A, “aff A” the affine hull of A, “lin A” the linear hull of µ A; “clµ(A) = A ” the µ−closure of A, “riµ A” the topological interior of A i with respect to clµ(aff A), “core A” the algebraic interior of A, “ A” the relative algebraic interior of A, and µ−icA := iA if aff A is µ−closed and µ−icA := ∅ otherwise, the relative algebraic interior of A with respect to clµ(aff A). When the topology µ is implicitly understood (such is the case when we deal with the strong topology of a normed space) the use of the µ−notation is avoided. For f,g : E → R := R∪{−∞, +∞} we set [f ≤ g] := {x ∈ E | f(x) ≤ g(x)}; the sets [f = g], [f<g], and [f >g] are defined similarly. Throughout this paper, if not otherwise explicitly mentioned, (X, k·k) is a non trivial (that is, X 6= {0}) normed space, X∗ is its topological dual endowed with the weak-star topology w∗, the topological dual of (X∗, w∗) is identified with X, the weak topology on X is denoted by w, and the strong topology on X is denoted by s. The closed unit ball of X is denoted by BX := {x ∈ X |kxk≤ 1}. The duality product of X × X∗ is denoted by hx, x∗i := x∗(x) =: c(x, x∗), for x ∈ X, x∗ ∈ X∗. As usual, for S ⊂ X, S⊥ := {x∗ ∈ X∗ | hx, x∗i = 0, for every x ∈ S}, ∗ ∗ ∗ ∗ ∗ ⊥ ∗ σS(x ) := supx∈Shx, x i, x ∈ X and for A ⊂ X , A := {x ∈ X | hx, x i =0, ∗ for every x ∈ A}, σA(x) = supx∗∈Ahx, x i, x ∈ X. To a multifunction T : X ⇒ X∗ we associate its graph: Graph T = {(x, x∗) ∈ X × X∗ | x∗ ∈ T (x)}, inverse: T −1 : X∗ ⇒ X, Graph T −1 = {(x∗, x) | (x, x∗) ∈ Graph T }, domain: D(T ) := {x ∈ X | T (x) 6= ∅} = PrX (Graph T ), and range: ∗ ∗ ∗ R(T ) := {x ∈ X | x ∈ T (x) for some x ∈ X} = PrX∗ (Graph T ). Here PrX ∗ ∗ and PrX∗ are the projections of X × X onto X and X , respectively. When no confusion can occur, T will be identified with Graph T . On X, we consider the following classes of functions and operators Λ(X) the class formed by proper convex functions f : X → R. Recall that f is proper if dom f := {x ∈ X | f(x) < ∞} is nonempty and f does not take the value −∞, 2 Γτ (X) the class of functions f ∈ Λ(X) that are τ–lower semicontinuous (τ–lsc for short); when the topology is implicitly understood we use the notation Γ(X), M(X) the class of non-empty monotone operators T : X ⇒ X∗. Recall that T : ∗ ∗ ∗ ∗ ∗ X ⇒ X is monotone if hx1 − x2, x1 − x2i≥ 0, for all (x2, x2), (x2, x2) ∈ T . M (X) the class of maximal monotone operators T : X ⇒ X∗. The maximality is understood in the sense of graph inclusion as subsets of X × X∗. To a proper function f : X → R and a topology τ on X we associate: • the epigraph of f: epi f := {(x, t) ∈ X × R | f(x) ≤ t}, • the convex hull of f: conv f : X → R, is the greatest convex function majorized by f, (conv f)(x) := inf{t ∈ R | (x, t) ∈ conv(epi f)}, x ∈ X, • the τ−lsc convex hull of f: clτ conv f : X → R, is the greatest τ–lsc convex function majorized by f, (clτ conv f)(x) := inf{t ∈ R | (x, t) ∈ clτ conv epi f} , x ∈ X, • the convex conjugate of f : X → R with respect to the dual system (X,X∗): f ∗ : X∗ → R, f ∗(x∗) := sup{hx, x∗i− f(x) | x ∈ X}, x∗ ∈ X∗. • the subdifferential of f at x ∈ X: ∂f(x) := {x∗ ∈ X∗ | hx′ − x, x∗i + f(x) ≤ f(x′), ∀x′ ∈ X} for x ∈ dom f; ∂f(x) := ∅ for x 6∈ dom f. Recall that NC = ∂IC is the normal cone of C ⊂ X, where IC is the indicator function of C ⊂ X defined by IC (x) := 0 for x ∈ C and IC (x) := ∞ for x ∈ X \ C. Let Z := X × X∗. It is known that (Z,s × w∗)∗ = Z via the coupling z · z′ := hx, x′∗i + hx′, x∗i , for z = (x, x∗), z′ = (x′, x′∗) ∈ Z. For a proper function f : Z → R all the above notions are defined similarly. The conjugate of f with respect to the natural dual system (Z,Z) induced by the previous coupling is given by f : Z → R, f (z) = sup{z · z′ − f(z′) | z′ ∈ Z}, and by the biconjugate formula, f = cls×w∗ conv f whenever f or cls×w∗ conv f is proper. We consider the following classes of functions on Z: C := C(Z) := {f ∈ Λ(Z) | f ≥ c}, R := R(Z) := Γs×w∗ (Z) ∩C(Z), D := D(Z) := {f ∈ R(Z) | f ≥ c}. It is known that [f = c] ∈M(X) for every f ∈C(Z) (see e.g. [24, Lemma 3.1]). To a multi-valued operator T : X ⇒ X∗ we associate the following functions: 3 • the Fitzpatrick function of T (introduced in [10]): ϕT : Z → R, ϕT := cT , where cT : Z → R, cT := c + ιT ; ∗ • ψT := cl s×w conv cT (first considered in [22]); ψT = ϕT = cT whenever ϕT or ψT is proper (for example when T ∈M(X) (see e.g. [23, Proposition 3.2])). Note that for every T ⊂ Z, [ϕT ≤ c] describes the set of all z ∈ Z that are monotonically related (m.r. for short) to T , that is, c(z − a) ≥ 0, for every a ∈ T . We call a multifunction T : X ⇒ X∗ unique if T admits a unique maximal monotone extension; representable in Z if T = [f = c], for some f ∈ R; in this case f is called a representative of T . We denote by RT the class of representatives of T ; this notion was first considered in this form in [22]; dual-representable if T = [f = c], for some f ∈ D; in this case f is called a d–representative of T and we denote by DT the class of d-representatives of T ; NI or of negative infimum type in Z if ϕT ≥ c in Z; demiclosed if Graph T is closed with respect to the strong×weak-star conver- ∗ ∗ gence of bounded nets in Z, that is, if xi → x0 strongly in X, xi → x0 ∗ ∗ ∗ ∗ weakly-star in X , (xi )i is (strongly) bounded in X , and {(xi, xi )}i ⊂ T ∗ then (x0, x0) ∈ T . strongly × weakly − star upper H − semicontinuous (s × w∗−usc for short) at x ∈ X if for every weak-star open set V ⊃ T x there exists a neighbor- hood U of x such that T (U) := ∪u∈U Tu ⊂ V (or equivalently for every net {xi}i∈I with xi → x, strongly in X, eventually T xi ⊂ V ).